The Agent Skills Directory

[SAFE]: The skill's primary function is to process and refine potentially vague or messy user input. It incorporates robust security mitigations, including mandatory clarifying questions and a requirement for explicit user confirmation (Step 7) before any refined prompt is executed. This design effectively prevents automated execution of malicious instructions embedded in user data.
[PROMPT_INJECTION]: The skill employs standard prompt engineering techniques such as persona assignment (Expert Persona) and instructional modifiers (Stakes Language). These are transparently used to enhance model performance and quality rather than to bypass safety protocols or override core agent constraints.
[COMMAND_EXECUTION]: The execution templates (e.g., in templates/work.md) recommend standard, transparent development commands like bun run lint, bun run type-check, and bun run build. These are documented as quality gates for a legitimate software development workflow and do not involve hidden or unauthorized shell execution.
[DATA_EXFILTRATION]: The skill reads project-level configuration (CLAUDE.md) to provide context for prompt refinement. No patterns of sensitive file access (e.g., credentials or SSH keys) or network operations to non-whitelisted domains were found.

prompt-refinement